Privacy policy

Summary

• Your recordings, transcripts, translations and summaries are stored on your device, encrypted. We never upload them for storage, backup or sync.
• By default, Hujam sends audio or text to our cloud service, Hujam AI, for processing. Hujam AI passes it to a specialist speech or language provider and returns the result to your device. We do not store your audio. A transcript result is held for at most 30 minutes so that we do not charge you twice. We do not use your content to train AI models.
• You can keep everything on your phone. Each capability (transcription, speaker identification, translation, summarization) has an on-device option. Choose it once in Settings and that capability never sends your content anywhere. It also works in airplane mode.
• No advertising, no data sales, and no analytics, tracking or crash-reporting SDKs. We have switched off the usage reporting of the one Google library the app uses (§ 5.8).
• You control deletion. You can erase all content on the device with one action. Deleting your account removes what we hold about you on our servers.
• Other people's voices. If you record someone, you are responsible for having the right to do so (§ 8).

1. Who we are and how to reach us

Hujam ("Hujam", "we", "us", "our") is provided by:

Umber AB Ribegatan 28, Stockholm 11

For any privacy question, request or complaint, email hello@umberab.com. We answer every message.

For the personal data described in this policy we are the data controller under the EU General Data Protection Regulation ("GDPR"), and the business under US state privacy laws.

EU representative (GDPR Art. 27): Not applicable: we are established in the EU at the address above

Data Protection Officer: we have not appointed one, because GDPR Art. 37 does not require it for our processing. Contact the address above.

2. What this policy covers

This policy covers:

• the Hujam Android app, including pre-release test builds distributed through Google Play testing tracks;
• Hujam AI, our cloud processing service, reached through our gateway at auth.snkdev.net;
• your Hujam account, subscriptions, and model downloads;
• messages you send us.

It does not cover services you reach by your own choice, such as Google Play, your Google account or Hugging Face. Their own privacy policies apply to them.

3. How Hujam works: on-device or Hujam AI

Hujam records audio, or imports audio you choose, transcribes it, can identify different speakers, and can translate and summarize the text. Each of these capabilities runs in one of two places. You choose where, separately for each capability, in Settings → Defaults or from the engine strip on the capability's own screen. The choice is saved, not asked on every run:

• On-device. Processing happens on your phone with a model stored on your phone. Nothing about your content leaves the device.
• Hujam AI (the default on a new install). Your audio or text is sent over an encrypted connection to our service. The result comes back and is stored on your phone.

Every transcript, translation, summary and speaker labelling carries a badge that shows whether it was produced on-device or by Hujam AI. If the option you chose cannot run (for example, you are offline or have no on-device model), Hujam tells you. It never switches to the other option silently. Changing where your data is processed always takes a tap from you.

4. Data that stays on your device

This data is created and stored only on your phone. We cannot see it, access it or recover it:

Android system backup is disabled for Hujam. Google Drive backup and device-to-device transfer do not copy this data.

Storing this data on your device is strictly necessary to provide the app you asked for (ePrivacy Directive Art. 5(3)). Hujam does not use cookies.

5. Data we process, why, and on what legal basis

This section explains each situation in which data leaves your device. For people in the EEA it also gives our legal basis under GDPR Art. 6. § 12 states how long we keep each item.

5.1 Hujam AI processing of your content

When: only when a capability is set to Hujam AI and you run it, or you turned on automatic processing for new recordings with that capability set to Hujam AI.

What is sent:

• Transcription and speaker identification: the audio of the recording, its declared language if you chose one, and your sign-in token.
• Translation and summarization: the text to be translated or summarized, the target language or summary style, and your sign-in token.

What happens to it:

1. Our gateway verifies your sign-in token, checks your plan and remaining allowance, and forwards your content to one of the processing providers listed in § 10. Your name, email address and account ID are not sent to the provider. The provider receives the content and the processing instructions, and nothing else about you.
2. The provider returns the result. We send it to your device.
3. Audio is not stored by us. It passes through our gateway's memory and is discarded when the request ends.
4. Short-lived result cache. Speaker identification and transcription of the same audio can be served by one provider call. To avoid charging you or us twice, we cache the returned transcript text and speaker turns (timestamps and labels such as "Speaker 1") for 30 minutes. The cache is keyed by a one-way cryptographic hash (SHA-256) of the audio. It stores no account identifier and no audio, and nothing in it can reconstruct the audio. After 30 minutes the entry can no longer be used, and it is permanently deleted within the following hour.

Purpose: to provide the transcription, speaker identification, translation or summary you asked for.

Legal basis: performance of our contract with you (Art. 6(1)(b)). If your content reveals special categories of personal data, see § 8.

We do not use your content to train, fine-tune or evaluate AI models. We do not read it, analyse it for any other purpose, or combine it with other data. Our processing providers are bound by contract to the same restrictions (§ 10).

This is optional. Hujam AI is only one of two ways to run each capability. To keep your content on your phone, select the on-device option (§ 13).

5.2 Your account

An account is optional. You can use the whole app during your free trial, and with a subscription, using on-device processing only, without one. An account is needed for Hujam AI and for downloading models from our server.

Purpose: to create and secure your account, sign you in, verify your email address, and link your subscription to you.

Legal basis: performance of contract (Art. 6(1)(b)). Keeping your account secure is also our legitimate interest (Art. 6(1)(f)).

Your content is never linked to your account or stored in it. No recordings, transcripts or other content are synced between devices.

If you sign in with Google, Google confirms your identity to us and shares your name, email address and profile-picture URL. Google's privacy policy governs what Google does.

5.3 Plan, allowance and usage counters

For each account, our gateway keeps usage counters: your account ID, the billing period, the capability, and how much you used (minutes of audio or a number of requests). It also keeps a short-lived copy of your subscription status (plan, tier, period dates) so it does not have to ask our authentication provider on every request. The counters contain no audio, no text and no IP address.

Purpose: to apply your plan's allowance, show you how much remains, and prevent abuse of the service.

Legal basis: performance of contract (Art. 6(1)(b)), and our legitimate interest in preventing fraud and abuse (Art. 6(1)(f)).

5.4 Model downloads

When you download an on-device model from us, our gateway checks your entitlement and gives your device a short-lived download link. We record your account ID, the model file requested, your plan tier and the time.

Purpose: to enforce entitlements, comply with the licence terms of the models we distribute, and detect abusive bulk downloading.

Legal basis: performance of contract (Art. 6(1)(b)), and our legitimate interests in licence compliance and abuse prevention (Art. 6(1)(f)).

Your device also downloads our model catalogue (the list of available models) and the Hujam AI configuration (such as the current allowance). No account data is sent with these requests.

5.5 Models from Hugging Face

If you enter a Hugging Face repository to download a model from, your device contacts Hugging Face directly. It sends the repository name and, if you added one, your Hugging Face access token. We are not part of that request, and Hugging Face acts as an independent controller under its own privacy policy. Your token is stored only on your device (§ 4).

5.6 Subscriptions and payments

Subscriptions are sold through Google Play Billing and managed for us by RevenueCat.

• Google processes your payment. We never receive your card or payment details. Google is responsible for the payment under its own terms.
• RevenueCat receives an app-specific customer identifier (after you sign in, your Hujam account ID), your purchase and renewal records from Google Play, and basic device and app information (such as OS version, app version and language). The RevenueCat library is started when the app opens so that your subscription status is always correct. RevenueCat reports your subscription status to our authentication provider, which stores it with your account (plan, whether active, tier, period dates).

Purpose: to sell and manage subscriptions, unlock paid features, and keep records required by law.

Legal basis: performance of contract (Art. 6(1)(b)). Keeping transaction records meets our legal obligations under tax and accounting law (Art. 6(1)(c)).

5.7 Messages you send us and pre-release testing

If you email us or send feedback, we process your email address, the contents of your message and anything you attach, in order to answer you and fix problems. Please do not attach recordings or transcripts unless we specifically need them to investigate an issue you reported.

While Hujam is in pre-release testing, you install it through a Google Play testing track. If we invited you by email, we keep that email address on our tester list to give you access. Feedback you send through Google Play's testing tools reaches us with the details Google attaches to it.

Legal basis: our legitimate interest in supporting users and improving a product under test (Art. 6(1)(f)). Where you signed up as a tester, performance of the testing arrangement (Art. 6(1)(b)).

5.8 Technical data sent by Google components

• On-device translation (Google ML Kit). The first time you translate between a pair of languages on-device, the app downloads the language packs from Google, so Google receives the request (your IP address and which pack). The text you translate is never sent: translation runs on your device. We have disabled the ML Kit library's usage reporting to Google. Google's handling of the download is covered by the Google Privacy Policy.
• Crash reports (Android vitals). If you have allowed your device to share usage and diagnostic data with Google, Google Play may give us aggregated crash and stability reports. These contain device model, Android version and technical traces of the failure, but not your content. We use them only to fix defects. You control this in your device's Google settings.

Legal basis: our legitimate interest in providing on-device translation and keeping the app stable (Art. 6(1)(f)).

5.9 Network data

Any connection to the internet exposes your device's IP address to the server it reaches. Our gateway and our providers necessarily process it to deliver responses and defend against attacks. We do not write your IP address to our application logs or databases, and we do not use it to locate or profile you.

Legal basis: our legitimate interest in operating a secure service (Art. 6(1)(f)).

5.10 Is providing this data required?

Providing personal data is voluntary. You can use Hujam without an account, fully on-device. Without an account you cannot use Hujam AI or download models from us. Without the relevant content you cannot use Hujam AI for that content. No law requires you to provide any data to us.

6. Legitimate interests

Where we rely on legitimate interests (Art. 6(1)(f)), those interests are: keeping the service and your account secure; preventing fraud and abuse of allowances; complying with model licences; answering you; and fixing defects. We have weighed these interests against your rights. The data involved is minimal, never includes your content, and matches what you would reasonably expect. You can object at any time (§ 14.1).

7. Consent

We do not rely on consent for our core processing. Before your first Hujam AI run of translation, summarization or speaker identification, the app shows you where your data will go. Where we do rely on your consent (§ 8), you may withdraw it at any time by switching that capability to on-device. Withdrawal does not affect processing that already happened.

8. Sensitive information and other people in your recordings

Sensitive content. Recordings can capture anything said aloud, including information about health, beliefs, sexuality or other matters that are "special categories of personal data" under GDPR Art. 9, or "sensitive" or "consumer health" data under US state laws. We do not ask for this information. We process it only if it happens to be in content you choose to send to Hujam AI, only to return the result you requested, and never for any other purpose. Where GDPR applies, we process special-category data on the basis of your explicit consent (Art. 9(2)(a)), which you give by choosing Hujam AI for that capability and running it. If a recording contains sensitive information, we recommend that you use on-device processing for it.

Voices are not used to identify anyone. Speaker identification divides a recording into turns labelled "Speaker 1", "Speaker 2" and so on. It does not determine who a speaker is. It does not create, store or compare voiceprints or other biometric identifiers, and it does not recognise the same person across recordings. The mathematical voice features computed to separate speakers exist only while that one recording is processed and are then discarded, on your device or by our provider. You can rename speakers yourself, and those names stay on your device.

Other people. Recording, and sending recordings to a cloud service, can be regulated by law. In many places, including several US states such as California, Florida, Illinois, Maryland, Massachusetts, Montana, New Hampshire, Pennsylvania and Washington, and under EU and national law, you may need everyone's consent before you record a conversation. You are responsible for having any consent and notice the law requires before you record or process other people's voices, and for how you use recordings you make. When you use Hujam purely for personal or household purposes, EU data protection law generally does not apply to your own use of those recordings. It still applies to our processing.

If you believe Hujam AI processed information about you through someone else's use of the app, contact us. Because we do not keep content, we generally hold nothing about you beyond the 30-minute cache described in § 5.1.

9. AI-generated results and automated decisions

Transcripts, speaker labels, translations and summaries are produced by machine-learning models and may contain errors. They are labelled with the engine that produced them. You can correct transcript text yourself, and corrected segments are marked as edited.

We do not make decisions about you based solely on automated processing, including profiling, that produce legal or similarly significant effects (GDPR Art. 22).

10. Who we share data with

We do not sell personal data and do not share it for advertising. We disclose it only to the providers below, who process it on our behalf and on our written instructions (GDPR Art. 28 processors; "service providers" or "processors" under US state laws). Each is contractually bound to use the data only to provide its service to us, to keep it confidential and secure, and not to use your content to train or improve its models.

Why several providers. For reliability, each Hujam AI capability has a primary provider and a backup. If the primary is unavailable, your request goes to the backup, which is also listed above. Your choice of Hujam AI versus on-device is never affected. We add providers to this list before any of your data is sent to them (§ 18).

The AI models these providers run for us include OpenAI's Whisper, Meta's Llama and OpenAI's gpt-oss. These are open models hosted by the providers above. Your data is not sent to OpenAI or Meta.

Independent controllers. Google (Google Play, Google sign-in, ML Kit) and Hugging Face (if you use it) process data under their own privacy policies, not on our behalf.

Legal disclosures. We may disclose data if the law requires it, for example under a valid court order. We will tell you unless the law forbids us. We may also disclose it to protect the rights, safety or property of users or others. Because we do not store your content, we generally have none to disclose.

Business transfers. If Hujam is merged, acquired or sold, personal data may pass to the new owner. The new owner must honour this policy, and we will tell you before your data becomes subject to a different policy.

11. International transfers

Some of our providers are based in, or process data in, the United States and other countries outside the EEA. When personal data of people in the EEA is transferred to a country without an EU adequacy decision, we rely on one of:

• the EU–US Data Privacy Framework, for providers certified under it;
• the European Commission's Standard Contractual Clauses (2021/914), together with an assessment of the destination's laws and supplementary measures such as encryption in transit and data minimisation.

To request a copy of the relevant safeguards, contact hello@umberab.com.

12. How long we keep data

13. Your controls in the app

• Keep content on your phone. Set any capability to on-device in Settings → Defaults. Once all capabilities are on-device, no content leaves your phone.
• Delete content. Delete any recording, transcript or project at any time.
• Erase everything on the device. Settings → Data & privacy contains a wipe action. It deletes every project, recording, transcript, summary, downloaded model, stored token and activity-log entry, and destroys the encryption keys.
• Delete your account. Account → Delete account removes your account and all related records on our servers and at RevenueCat: email, profile, subscription status, usage counters and download log. This cannot be undone. You can also request deletion without the app at https://hujam.umberab.com/delete-account or by emailing hello@umberab.com.
• Export your content. Export projects and transcripts to a file you control (data portability).
• Correct your data. Edit your display name in Account. Edit transcript text in the player.
• Cancel your subscription in Google Play. Uninstalling the app does not cancel it.

14. Your privacy rights

We honour the rights below for all users, wherever you live. There is no charge. Most of your data is under your direct control on your device (§ 13). For data on our servers, email hello@umberab.com or use the in-app controls. We will not discriminate against you for exercising any right.

14.1 EEA (and UK and Switzerland)

Under GDPR (and the UK GDPR and Swiss FADP) you have the right to:

• access your personal data and receive a copy (Art. 15);
• rectify inaccurate data (Art. 16);
• erase your data (Art. 17);
• restrict processing (Art. 18);
• data portability (Art. 20);
• object to processing based on legitimate interests (Art. 21);
• withdraw consent at any time, without affecting earlier processing (Art. 7(3));
• lodge a complaint with a supervisory authority, in particular in the country where you live or work, or where an alleged infringement took place (Art. 77). The EU authorities are listed at https://www.edpb.europa.eu/about-edpb/about-edpb/members_en. Our lead authority is the Swedish Authority for Privacy Protection (Integritetsskyddsmyndigheten, IMY), https://www.imy.se. We would appreciate the chance to address your concern first.

We respond within one month. For complex requests the law allows two further months, and we will tell you if we need them.

14.2 United States

Residents of California and of other states with comprehensive privacy laws (including Colorado, Connecticut, Delaware, Indiana, Iowa, Kentucky, Maryland, Minnesota, Montana, Nebraska, New Hampshire, New Jersey, Oregon, Rhode Island, Tennessee, Texas, Utah and Virginia) have, depending on the state, the right to:

• know / access the personal data we hold and how we use and disclose it, and receive it in a portable format;
• correct inaccurate personal data;
• delete personal data;
• opt out of the sale or sharing of personal data, targeted advertising, and profiling with legal or similarly significant effects. We do none of these. We also honour Global Privacy Control signals where they reach us;
• limit the use of sensitive personal data. We use it only for purposes that the law allows without an opt-out, such as providing the service you requested and securing your account;
• appeal our decision on your request. Reply to our decision with "Appeal" in the subject line. We answer within the time your state's law allows (generally 45–60 days). If you disagree with the result, you can contact your state Attorney General.

We respond within 45 days, extendable once by 45 days where the law allows. We verify your request by confirming control of your account's email address. An authorized agent may submit a request on your behalf with your signed permission, and we may ask you to confirm your identity directly.

California notice at collection (CCPA/CPRA). In the past 12 months we have collected the following categories of personal information. Every category is used only for the purposes in § 5 and kept as stated in § 12. We sold or shared none of it (as those terms are defined in the CCPA) and have no actual knowledge of selling or sharing data of consumers under 16.

We do not collect precise geolocation, biometric identifiers, contacts, or information for advertising, and we do not draw inferences to build a profile about you.

Washington, Nevada and similar consumer health data laws. We do not seek consumer health data. If a recording you send to Hujam AI contains health information, we process it only to provide the service you requested (§§ 5.1, 8). We never sell it, and it is covered by the same rights, deletion and provider terms as other content. You can ask us to confirm what we hold and delete it (in practice, at most the 30-minute cache).

15. Children

Hujam is not directed to children. You must be at least 16 to create an account or subscribe. We do not knowingly collect personal data from anyone under that age, or from any child under 13 (US Children's Online Privacy Protection Act). If you believe a child has given us personal data, contact hello@umberab.com and we will delete it.

16. Security

• Content on your device is encrypted at rest. The keys are held in the Android Keystore and never leave the device.
• All traffic between the app, our gateway and our providers uses TLS encryption.
• Access to Hujam AI requires a valid, signed sign-in token. Administrative functions require a separate operator credential. Provider credentials are kept as encrypted secrets and never shipped in the app.
• We collect the minimum data needed, and never store your content on our servers except the 30-minute cache.

No system is perfectly secure. Because your content is stored on your phone, the security of your device (screen lock and system updates) is an important part of protecting it. If a personal-data breach is likely to put your rights at risk, we will notify the competent supervisory authority within 72 hours and inform you without undue delay where the law requires.

17. Android permissions

Hujam does not request access to your location, contacts, calendar, camera, photos, call logs, SMS or list of installed apps. You choose each file to import or export yourself, through Android's file picker.

18. Changes to this policy

We update this policy whenever our data handling changes, and we change the "Last updated" date above. Before a material change takes effect, such as a new processing provider, a new kind of data, or a new purpose, we notify you in the app and, if you have an account, by email. Previous versions are available on request.

19. Contact

Umber AB Ribegatan 28, Stockholm hello@umberab.com